It is well established that hypoxia potently stimulates tumor angiogenesis by activating hypoxia inducible factor-1 (HIF-1)–induced proangiogenic factors, such as vascular endothelial growth factor. However, very little is known about the role of hypoxia in incipient angiogenesis in avascular tumors during their early stages of growth. To noninvasively investigate the functional significance of hypoxia and HIF-1 activation in incipient tumor angiogenesis, we genetically engineered HCT116 human colon carcinoma cells and 4T1 mouse mammary carcinoma cells with constitutively expressed red fluorescence protein as a tumor marker and green fluorescence protein (GFP) as a reporter for hypoxia and HIF-1 activation. The accuracy of GFP fluorescence in reporting hypoxia was confirmed by flow cytometry analysis and by immunohistochemical comparison with pimonidazole, a well-established hypoxia marker drug. Mouse dorsal skin-fold window chambers showed that incipient angiogenesis preceded a detectable level of hypoxia. The detectable levels of hypoxia were spatially and temporally related with more intensive secondary angiogenesis following the initial onset of new vessel formation. Selective killing of hypoxic cells by tirapazamine efficiently eliminated or delayed the detection of hypoxic cells, but it did not significantly delay the onset of incipient angiogenesis. These findings provide the first in vivo evidence that incipient tumor angiogenesis may not depend on hypoxia or HIF-1 activation. This is in contrast to the clear role of hypoxia in driving angiogenesis once initial tumor microvessel formation has occurred.
Although tumors vary tremendously in their genetic alterations, invasiveness, and metastatic potential, continued tumor growth and metastasis depend on angiogenesis—the formation of new blood vessels from preexisting vasculature (1). Currently, very little is known about how much incipient tumor angiogenesis is attributable to hypoxia, which is one of the most common microenvironmental factors in established tumors. The hypoxic response is mediated to a large extent by hypoxia inducible factor-1 (HIF-1), a basic helix-loop-helix transcription factor heterodimerized by HIF-1α and HIF-1β subunits; HIF-1β is constitutively expressed, whereas HIF-1α is posttranslationally regulated by ubiquitination-mediated proteolysis (2, 3). Many studies have shown that hypoxia and HIF-1 activation stimulate angiogenesis in mature tumors by up-regulating multiple proangiogenic factors, including vascular endothelial growth factor (VEGF; ref. 4). For example, overexpression of HIF-1 in HCT116 tumor cells with a vector encoding HIF-1α markedly promotes VEGF expression and angiogenesis (5). Peptide blockade of HIF-1α degradation is also effective in stimulating angiogenesis (6). Although it is not surprising for one to speculate that the hypoxia response may play a significant role in the induction of incipient angiogenesis by up-regulating hypoxia-induced proangiogenic factors, this speculation has not been tested in tumors before the onset of angiogenesis in vivo. In this study, we noninvasively investigated whether hypoxia and HIF-1 activation significantly affect incipient angiogenesis by creating tumor cells with red fluorescence protein (RFP) as the marker for tumor visualization and green fluorescence protein (GFP) as the reporter for hypoxia and HIF-1 activation. Allograft and xenograft tumors in mouse dorsal skin-fold window chambers revealed that incipient tumor angiogenesis preceded detectable levels of hypoxia. To further assess the role of hypoxia in these early phases of angiogenesis, we treated mice with tirapazamine, a hypoxia-selective cytotoxin, at the time of tumor inoculation. We then continued tirapazamine treatments daily until the end point of the experiment. Tirapazamine treatment efficiently suppressed hypoxic cells but did not significantly delay the onset of incipient tumor angiogenesis. These results suggest that a hypoxic response may not be required for incipient tumor angiogenesis. These findings also provide new rationales to optimize current hypoxia/HIF-1 targeting therapies.
Materials and Methods
DNA constructs. To construct a hypoxia-responsive fluorescence protein reporter, an artificial hypoxia-responsive promoter (HRP), which consists of five copies of a VEGF-derived hypoxia-responsive element with a minimal cytomegalovirus (CMV) promoter, was excised from pHRP-Luc+ plasmid (7) by XhoI/HindIII and ligated into pEGFP-1 (Clontech, Palo Alto, CA) to control the expression of an enhanced GFP (EGFP) gene (8). Subsequently, the HRP-EGFP cassette was transferred into pSIR (Clontech), a retroviral plasmid that inactivates the 5′ long terminal repeat (LTR) promoter after reverse transcription. This vector was selected because the activities of the hypoxia-responsive promoter are not affected by the viral 5′ LTR promoter. To obtain an RFP reporter construct, the dsRed1 gene from pDsRed1 (Clontech) was cloned into the plasmid pLPCX (Clontech) and driven by a human CMV promoter for retroviral packaging.
Cell culture. Human HCT116 colon cancer cells were cultured with McCoy's 5A medium (Life Technologies, Carlsbad, CA). Mouse 4T1 mammary carcinoma cells were cultured with DMEM (Life Technologies). Both media were supplemented with 10% (v/v) fetal bovine serum (FBS, HyClone, Logan, UT) and 1% (v/v) antibiotics-antimycotics (Life Technologies).
Establishment of double-fluorescent reporter cells. Retrovirus vectors encoding the above HRP-GFP and CMV-RFP reporters were produced after transfecting the packaging cell line ψX-ampho with the corresponding retroviral plasmid by LipofectAMINE (Invitrogen, Carlsbad, CA). The supernatant media of the transfected ψX-ampho cells were used to infect HCT116 and 4T1 cells. Three days after infection, clonal cell populations that had stably integrated the retrovirus vectors were first selected by G418 (1 mg/mL) for the HRP-GFP reporter and puromycin (3 μg/mL) for the CMV-RFP reporter. The resistant cells were then subjected to 0.5% hypoxia and selected by fluorescence-activated cell sorting. Cells with constitutive RFP expression and hypoxia-induced GFP expression were selected.
Flow cytometry analysis. Hypoxia reporter (GFP)–transduced HCT116 and 4T1 cells were cultured in 25 mmol/L HEPES-buffered media under hypoxia (0.5% O2, 5%CO2, N2 balanced) versus normoxia (21% O2, 5% CO2) for 0, 6, 9, 12, 18, 24, 36, 48, and 72 hours. The mean fluorescence intensity of GFP in 10,000 cells was quantified by FACSCalibur flow cytometer (Becton Dickinson, Franklin Lakes, NJ) at each time point. Parental HCT116 and 4T1 cells were analyzed in parallel as negative controls.
Immunohistochemistry. To test the hypoxia-induced GFP expression in vitro, both HCT116 and 4T1 reporter cells were cultured under the above hypoxic or normoxic conditions for 11 hours. Cells were then stained with Hoechst 33342 (Sigma, St. Louis, MO) and imaged by a Hamamatsu C2741-08 CCD camera (Hamamatsu Photonics, Hamamatsu City, Japan). To analyze the distribution of the perfusion marker Hoechst 33342, the hypoxia marker pimonidazole, and the GFP expression in tumor tissue, we established tumors from the above two reporter cell lines (HCT116 in NCr/nu mice and 4T1 in BALB/c mice). When tumor size reached 5 to 7 mm in diameter, pimonidazole hydrochloride (70 mg/kg, i.p., Chemicon International, Inc., Temecula, CA) and Hoechst 33342 (20 mg/kg, i.v.) were injected 3 hours and 2 minutes, respectively, before sacrifice. Cryosections (10 μm for Hoechst 33342 and 40 μm for pimonidazole staining) were fixed with 2% paraformaldehyde (4°C, 30 minutes), blocked, and then stained with rabbit antipimonidazole serum (1:200, 1 hour at room temperature) and rhodamine-red X–conjugated donkey anti-rabbit IgG antibody (1:100; 1 hour at room temperature; Jackson ImmunoResearch Laboratories, West Grove, PA). Slides were imaged using a Zeiss LSM510 confocal microscope. The positive areas of GFP and pimonidazole staining were quantified by LSM510 software; 144 fields and 136 fields were analyzed for HCT116 and 4T1 tumors, respectively (five to seven random fields per slide, five slides per tumor, and five tumors per cell line). Staining without primary antibody was used as the negative control.
Mouse dorsal skin-fold window chamber and intravital microscopy. Murine window chamber surgery and intravital microscopy were done as described (9, 10). Double fluorescent (RFP-GFP) reporter cells were cultured in ambient conditions with low confluence (<50%) and high viability (>99%). No cellular GFP fluorescence was seen before inoculation. HCT116 (1,000-1,500) or 4T1 (500-800) cells were inoculated into window chambers on day 1 (24 hours after window chamber surgery). Each animal was injected with tirapazamine (10.8 mg/kg/d, i.p.) with 0.1 mL saline or saline alone (0.1 mL/d, i.p.) at the time of tumor inoculation and continued for 10 days. Tumor area (as indicated by RFP fluorescence), tumor hypoxia (as indicated by GFP fluorescence), and vasculature were observed daily using a Zeiss intravital fluorescence microscope. Tumor vascular length density was used as an indicator of tumor angiogenesis. It was calculated using the following formula: vascular length density (mm/mm2) = total vascular length in tumor region / area of tumor region. The area of tumor region is defined as RFP+ area plus a 200-μm-wide circumference outside the RFP+ region. Vascular length densities at each time point were normalized to value obtained on day 1. The relative angiogenic rate at each time point was calculated as follows: (vascular length density − vascular length density of the prior day) / vascular length density of the prior day.
ELISA. HCT116 and 4T1 cells were plated in a six-well plate (1 × 106 cells/well) with DMEM (1% FBS) and cultured under the above normoxic or hypoxic conditions for 24 hours. VEGF protein levels in the cell culture medium were quantified by a Quantikine M human and mouse VEGF ELISA kit (R&D Systems, Minneapolis, MN).
Statistical analysis. A two-sided two-sample Wilcoxon test and a Kaplan-Meier analysis were done as described in Results. A log-rank test was used to compare tirapazamine versus saline treatment for time to angiogenesis and time to hypoxia. A Student's t test was used to analyze the results of relative vascular length density, relative angiogenic rate, and VEGF ELISA. P < 0.05 was considered significant.
Results and Discussion
Validation of the efficacy of the green fluorescence protein reporter in response to hypoxia in vitro. To noninvasively visualize tumor hypoxia and HIF-1 activation, we first transduced HCT116 and 4T1 cells, which have normal hypoxia-inducible HIF-1 activity, with a GFP reporter gene driven by a hypoxia-responsive promoter. To test the induction of GFP expression under hypoxic conditions, we cultured both reporter cell lines under hypoxic and normoxic conditions for 0 to 72 hours. Flow cytometry analysis revealed a significant linear increase in GFP mean fluorescence intensity as a function of hypoxia exposure time. In contrast, both cells cultured under normoxic conditions maintained only background fluorescence intensity (Fig. 1A-B). The mean fluorescence intensity of hypoxia-treated cells was significantly higher than that of normoxic cells (P < 0.05, n = 5). Hypoxia-induced GFP fluorescence in both cell lines was also detectable by fluorescence microscopy (Fig. 1C-D). These results show that both transduced cell lines can report hypoxia by increasing GFP fluorescence.
Estimate the efficacy of the green fluorescence protein reporter in visualizing hypoxia in tumor tissue. I.v. administration of the fluorescent dye, Hoechst 33342, has been widely used to highlight perivascular regions and indicate tissue perfusion. Previous studies showed that tumor hypoxia mainly exists in tissues with insufficient perfusion as visualized by absent or weak Hoechst 33342 fluorescence (11). To test whether the GFP reporter gene in the transduced HCT116 and 4T1 tumors was predominantly expressed in insufficiently perfused tissue regions where hypoxia usually exists, we i.v. injected Hoechst 33342 and compared the spatial distributions of GFP fluorescence and Hoechst 33342 labeling. Fluorescent images of cryosections revealed that intensive GFP fluorescence was predominantly restricted to regions with poor perfusion or distant from perfused vessels. In contrast, little or no GFP fluorescence was found in perfused regions as revealed by the scarce overlap between GFP fluorescence and Hoechst 33342 (Fig. 2A). This complementary distribution between GFP fluorescence and Hoechst 33342 labeling in HCT116 and 4T1 tumors indicates that the GFP reporter gene is specifically expressed in insufficiently perfused tumor tissues, where hypoxia commonly occurs.
To directly evaluate whether GFP can be used as a hypoxia-responsive reporter, we compared spatial distributions of GFP fluorescence and pimonidazole staining in HCT116 and 4T1 tumors. Pimonidazole is a well-characterized exogenous hypoxia marker (12). This drug forms protein adducts under hypoxic conditions that can be visualized by immunohistochemical staining. Therefore, comparison of distributions of GFP fluorescence and pimonidazole provides direct evidence of the efficiency of the GFP reporter in detecting tumor hypoxia. The majority of GFP+ tumor regions were either colocalized with or surrounded by pimonidazole staining (Fig. 2B). Interestingly, pimonidazole staining was absent in some central areas of GFP+ regions. This might reflect minor differences between the expression of the endogenous GFP reporter gene in a hypoxic region and the diffusion of the exogenous pimonidazole to the same hypoxic region. Quantitative analysis of 144 fields of five HCT116 reporter tumors and 136 fields of five 4T1 reporter tumors revealed that 98.0 ± 0.4% GFP+ areas in HCT116 tumors and 94.2 ± 1.0% GFP+ areas in 4T1 tumors were colocalized with or surrounded by pimonidazole staining. The GFP-pimonidazole concordant area was significantly higher than the nonconcordant area in all HCT116 and 4T1 tumors (n = 5; t test, P < 0.001). The above results confirm that the expression of the GFP reporter gene is suitable to visualize in vivo tumor hypoxia.
Spatiotemporal interactions between incipient tumor angiogenesis and hypoxic response. To visualize all tumor cells irrespective of hypoxia response, we transduced the above two hypoxia-GFP reporter cell lines with a constitutively expressed RFP gene. Serial monitoring in dorsal skin-fold window chambers allowed for simultaneous observation of tumor growth, hypoxia, and angiogenesis. We observed that incipient tumor angiogenesis preceded onset of detectable levels of hypoxia in HCT116 tumors (Fig. 3A). Tiny new vascular sprouts and branches grew from preexisting vessels at the periphery of HCT116 cell clusters 1 to 2 days before GFP fluorescence appeared. This finding suggests that some angiogenic activity already existed before hypoxia developed in those avascular HCT116 tumors. However, once the GFP fluorescence was observed, it was accompanied with or followed by a second wave of more robust tumor angiogenesis. Spatially, this secondary intensive angiogenesis occurred in the tumor periphery as well as inside the tumor. Often, vascular extension was observed toward the GFP+ tumor regions (Fig. 3A, white dashed circle). The 4T1 window chamber tumors showed similar results: Incipient angiogenesis preceded detection of hypoxia. A second wave of more intensive angiogenesis correlated with onset of hypoxia (Fig. 3B, white dashed circle).
To quantitatively analyze tumor angiogenesis, we measured tumor vascular length density at multiple time points before and after hypoxia. More intensive angiogenesis was observed on the day when hypoxia was first detected in HCT116 and 4T1 tumors, compared with the angiogenesis that occurred before this time (Fig. 3C). This indicates that there might be some relationship between hypoxia and the secondary more robust angiogenesis. To distinguish whether this more robust angiogenesis occurred just coincidentally with hypoxia or hypoxia could increase angiogenic activity, we analyzed the relative angiogenic rate at multiple time points before and after hypoxia was first detected. The maximal relative angiogenic rate was reached on the day that hypoxia was first detected in both tumor types. The rate was significantly higher in 4T1 tumors and the same trend was observed with HCT116. It suggests that hypoxia or HIF-1 activation likely contributes to the secondary wave of angiogenesis in 4T1 tumors (Fig. 3D).
Confirmation of hypoxia-independent incipient angiogenesis by selective elimination of hypoxic tumor cells. Although the above results indicate that incipient tumor angiogenesis occurs independently of hypoxia, it is possible that an undetectable level of hypoxia, as reflected by the GFP reporter, might induce incipient angiogenesis. To test this hypothesis, we selectively killed hypoxic cells with tirapazamine—a highly effective, hypoxia-selective cytotoxin (13)—and determined whether elimination of hypoxic cells can delay incipient angiogenesis. We compared the time required for incipient angiogenesis in tirapazamine versus saline-treated window chamber tumors. In saline-treated HCT116 tumors, incipient angiogenesis started before the appearance of GFP fluorescence (Fig. 4A) in a manner identical to the untreated tumors above (Fig. 3A). Tirapazamine treatment resulted in complete suppression of hypoxia-induced GFP expression in three of eight HCT116 tumors (Fig. 4B) and caused a statistically significant delay in the onset of GFP fluorescence (Fig. 4C). However, tirapazamine treatment did not delay the onset of incipient angiogenesis (Fig. 4D). Tirapazamine-treated 4T1 tumors showed similar results (Fig. 2E-H). The finding that selective killing of hypoxic cells does not affect the induction of incipient angiogenesis is directly evidence that hypoxia may not be a significant driving force for incipient tumor angiogenesis.
Hypoxia-independent vascular endothelial growth factor secretion by HCT116 and 4T1 tumor cells. Although hypoxia can induce VEGF expression by HIF-1 activation, it has recently been reported that some tumor cells can secrete VEGF independent of hypoxia or HIF-1 signaling (14). To test whether there could be hypoxia-independent VEGF secretion by HCT116 and 4T1 cells, we cultured both tumor cells under normoxic and hypoxic conditions. ELISA assays showed that hypoxia significantly increased VEGF secretion in both cell lines (n = 6; t test, P < 0.001). More importantly, both cell lines secreted low levels of VEGF even under normoxic conditions (Fig. 4I). Our finding of hypoxia-independent VEGF secretion by HCT116 cells is consistent with a recent report showing the mutant K-ras oncogene in HCT116 cells results in HIF-1–independent angiogenic activities (15). Although there could be other non–hypoxia-regulated proangiogenic factors inducing incipient tumor angiogenesis, VEGF secretion caused by genetic mutations or other factors may participate in this process. Consistent with this concept is our previous observation that neutralizing VEGF by a soluble VEGF receptor significantly delays incipient angiogenesis in 4T1 window chamber tumors (10).
Although hypoxia is the main factor that activates HIF-1, genetic mutations in some HIF-1 regulators, such as pVHL and PTEN, can also activate HIF-1 independent of hypoxia (16). Therefore, expression of HIF-1 reporter in the cell lines with those genetic mutations may not highly relate to hypoxia. However, it does not seem that the majority of the GFP reporter expression in the HCT116 and 4T1 cells in this study reflect those genetic mutations because of the following reasons: (a) in contrast to the constitutive non–hypoxia-induced HIF-1 activity in cells with pVHL mutations, neither the HCT116 nor the 4T1 cell lines show constitutive HIF-1 activity (17), which is supported by a lack of GFP expression under normoxic conditions; (b) pimonidazole staining confirmed the high concordance between GFP expression and hypoxia in both tumors; (c) in any case, the reporter is a robust indicator of HIF-1 activation. When GFP is not observed, it is clear that the HIF-1 pathway is not activated by any mechanism.
The spatiotemporal absence of GFP expression before or concordant with incipient angiogenesis and the ineffectiveness of tirapazamine in delaying incipient angiogenesis reveal a two-stage model of angiogenesis. The first phase clearly occurs independent of hypoxia or HIF-1 activation. In contrast, the hypoxia response appears to stimulate more intensive tumor angiogenesis at a later stage. The secretion of low-level VEGF by both cells under normoxic conditions suggests that alternative proangiogenic signaling(s) independent of hypoxia and HIF-1 activation may contribute to incipient angiogenesis. Identifying such proangiogenic signaling pathways may provide new insights and therapeutic targets to inhibit incipient angiogenesis.
The above findings also have clinical implications. Therapeutic reagents targeting tumor hypoxia and HIF-1 are actively pursued as a booming new category of drugs that may convert hypoxia from therapeutic obstacles into tumor-targeting advantages (18). One promise of these therapies is that they could significantly inhibit tumor angiogenesis by blocking HIF-1 signaling (16, 19). This study showed that suppression of the hypoxia response may not be sufficient to prevent or delay the progress of incipient tumor angiogenesis. Although those drugs may still be promising as antiangiogenic agents, it will be intriguing to explore whether additive or synergetic therapeutic benefits could be achieved by combining antiangiogenic therapy with the hypoxia/HIF-1 targeting therapies, potentially combined with radiotherapy or chemotherapy.
Note: B.J. Moeller, D. Yu, and Y. Zhao contributed equally to this work.
Grant support: Department of Defense Breast Cancer Research Program fellowship DAMD17-02-1-0368 (Y. Cao); NIH grant CA40355 and a National Cancer Institute breast cancer research Specialized Programs of Research Excellence grant (M.W. Dewhirst); NIH grants EB001882, CA81512, and a Komen Foundation breast cancer research grant (C-Y. Li).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Dr. J. Martin Brown (Department of Radiation Oncology, Stanford University School of Medicine, Stanford, CA) for providing the hypoxia responsive pHRP-Luc+ plasmid; Dr. Garry P. Nolan (Department of Microbiology and Immunology, Stanford University School of Medicine, Stanford, CA) for providing ψX-ampho cell line; Drs. William R. Wilson and Michael P. Hay (Auckland Cancer Society Research Center, the University of Auckland Faculty of Medical & Health Sciences, Auckland, New Zealand) for providing tirapazamine; Dr. James Raleigh (Department of Radiation Oncology, University of North Carolina School of Medicine, Chapel Hill, NC) for providing rabbit antipimonidazole serum; Dr. Thusitha Dissanayake, Lynn Martinek, and Mike Cook of the Flowcytometry Shared Resources of Duke University Comprehensive Cancer Center for help in flowcytometry analysis; Dr. Fan Yuan and Dr. Sheng Tong for assistance in confocal microscopy; and Jie Li for technical help in immunostaining.